ByteDance is tightening the reins on its large model system with a stronger strap. On October 13, its AI model department Seed issued an internal notice, deciding to establish a new department specifically responsible for managing the safety risks of large models. This new department is no ordinary one—it will report directly to Wu Yonghui, the "number one" in ByteDance's large model efforts, bypassing intermediate levels, highlighting the importance of this matter within the company.

The new department is named "Seed Model - Alignment & Preparedness," led by Yan Lin. Yan Lin is one of the core executives in ByteDance's Seed team, who has been in charge of the "alignment" of the Doubao large model for several years. According to ByteDance's external explanation, the purpose of establishing this department is to invest more heavily in safety and alignment. In simple terms, "alignment" means ensuring that the model always acts in line with human intentions and safety boundaries, preventing it from going off track. In terms of product application, you can directly feel its presence in the compliance, quality of answers, and even the value orientation of chatbots like Doubao.